Preamble

Production of µBq/m³ Level Low-²²⁶Ra Ultrapure Water for the Jiangmen Underground Neutrino Observatory

Xi-Ting Xu¹, Cong Guo²,³,⁴,†, Jin-Chang Liu²,³,⁴, Yong-Peng Zhang²,³,⁴, Xiao-Hua Liang⁵, Xin-Jian Wen¹,‡, Chang-Gen Yang²,³,⁴, Quan Tang⁶,§, Li-Dan Lv⁶, Yuan-Hao Niu¹, and Bin Xiao⁷

¹Institute of Theoretical Physics, Shanxi University, Taiyuan, 030006, China
²Experimental Physics Division, Institute of High Energy Physics, Chinese Academy of Sciences, Beijing, 100049, China
³School of Physics, University of Chinese Academy of Sciences, Beijing, 100049, China
⁴State Key Laboratory of Particle Detection and Electronics, Beijing, 100049, China
⁵Astro-particle Physics Division, Institute of High Energy Physics, Chinese Academy of Sciences, Beijing, 100049, China
⁶School of Nuclear Science and Technology, University of South China, Hengyang, 421001, China
⁷School of Mechanics and Optoelectronic Physics, Anhui University of Science and Technology, Huainan, 232001, China

The Jiangmen Underground Neutrino Observatory (JUNO), a 20-kiloton low-background liquid scintillator detector, was primarily designed to determine the neutrino mass ordering. JUNO requires ultrapure water (UPW) with a ²²⁶Ra concentration below 50 µBq/m³ due to direct liquid-liquid contact between the liquid scintillator (LS) and UPW during detector filling. To meet this stringent requirement, we developed a highly sensitive measurement system capable of detecting ²²⁶Ra at 3.9 µBq/m³ and optimized the 100 t/h UPW production process. By integrating selective ion-exchange resin with membrane separation technologies, we consistently produced UPW with a ²²⁶Ra concentration below 4 µBq/m³, exceeding JUNO's specifications and establishing a world-leading benchmark. This paper describes the design and implementation of JUNO's UPW system and the highly sensitive ²²⁶Ra measurement system, along with a systematic evaluation of ²²⁶Ra removal efficiency across purification stages and final water quality validation.

Keywords: JUNO, Ultra Pure Water, ²²⁶Ra, Reverse Osmosis, Resin

Introduction

The Jiangmen Underground Neutrino Observatory (JUNO) [1, 2] is a multipurpose experiment designed primarily to determine the neutrino mass ordering and precisely measure neutrino oscillation parameters by detecting reactor antineutrinos. With 20 kilotons of low-background liquid scintillator and an unprecedented energy resolution of 3% at 1 MeV, JUNO is the largest LS-based underground neutrino observatory, capable of addressing numerous important topics in astroparticle physics. JUNO's extensive physics program includes supernova neutrinos, atmospheric neutrinos, solar neutrinos, geoneutrinos, and searches for new physics [1, 2]. Located in an underground laboratory with 700 m of overburden, the muon flux at the JUNO site is approximately 4 mHz/m² with a mean energy of 207 GeV [3]. The primary detection channel for reactor antineutrinos in JUNO is inverse beta decay on protons. JUNO will detect only approximately 60 reactor antineutrino events per day, making strict control of radioactive background essential [4].

[FIGURE:1] shows a schematic drawing of the JUNO detector, which comprises a Central Detector (CD), a Water Cherenkov Detector (WCD), and a Top Tracker (TT) detector. The CD contains 20 kilotons of ultrapure LS within a spherical acrylic vessel and is submerged in the WCD, which contains 40 kilotons of ultrapure water and is equipped with 2400 20-inch MicroChannel Plate Photomultiplier Tubes (MCP-PMTs). The WCD provides sufficient water shielding to protect the CD from surrounding radioactivity and also serves as a cosmic muon veto. For background and safety considerations, the CD and WCD must initially be filled simultaneously with UPW. Subsequently, the UPW in the CD is replaced with LS at a production speed of 7 m³/h. During this LS replacement process, UPW and LS will be in direct contact. Since this is the final stage before detector operation, strict control of radionuclides in the water is essential to prevent UPW from contaminating the LS. One of the most challenging tasks is controlling the ²²⁶Ra concentration in the water to below 50 µBq/m³ [2].

In the field of low-background water purification and ²²⁶Ra detection, both the Sudbury Neutrino Observatory (SNO) [5] and Super-Kamiokande (Super-K) [6] experiments have achieved internationally recognized advances. SNO developed a measurement technique using manganese oxide-coated beads, reaching a ²²⁶Ra detection sensitivity of approximately 10 µBq/m³. The initial ²²⁶Ra concentration in SNO's UPW was approximately 50 µBq/m³, later reduced to 10–20 µBq/m³ through recirculation and purification [7]. Super-K adopted SNO's methodology and achieved a similar sensitivity of approximately 10 µBq/m³, though with slightly higher ²²⁶Ra levels in its UPW: initially 61 ± 13 µBq/m³, falling to 32 ± 7 µBq/m³ after recirculation [8]. In terms of production scale, Super-K's water system capacity is 30 t/h [8], notably larger than SNO's 9 t/h system [5].

This paper is structured as follows. Section II describes the UPW system of the WCD. Section III details the highly sensitive ²²⁶Ra measurement system with 3.9 µBq/m³ sensitivity. The measurement results of ²²⁶Ra concentration after each purification stage and in the final product water are presented in Section IV. Potential upgrades to the system for enhanced ²²⁶Ra removal are discussed in Section V, and a summary of the work is provided in Section VI.

UPW System

The JUNO experiment employs a 100 t/h UPW system to provide 40 kilotons of UPW for the WCD. This system produces water meeting the following specifications: resistivity greater than 18.2 MΩ·cm, ²²²Rn concentration below 1 × 10⁴ µBq/m³, and minimum temperature of 17 °C, ensuring high muon detection efficiency in the WCD, negligible radioactive contribution to the CD, and thermal stability for the CD [1]. Furthermore, for the initial filling of the acrylic vessel, this system must deliver UPW at 50 t/h with a ²²⁶Ra concentration below 50 µBq/m³ [2].

The JUNO 100 t/h UPW system utilizes tap water as feedwater to produce UPW compliant with China's national electronic grade EW-I standard [9]. As shown in [FIGURE:2], the system comprises aboveground and underground sections. The aboveground treatment system consists of sequential units: bag filters (5 µm pore size), multi-media filters (stratified quartz sand, activated carbon, anthracite), activated carbon filters (coconut shell-derived carbon), softening resin columns (special ²²⁶Ra removal resin from a Chinese manufacturer), cartridge filters (5 µm pore size), and reverse osmosis (RO) membranes.

Pretreatment begins with mechanical filtration via bag filters for bulk particulate removal, followed by multi-media filtration for suspended solids reduction. Subsequent activated carbon adsorption removes residual chlorine, organic compounds, ammonia-nitrogen species, nitrites, trace heavy metals, and microbial contaminants. Ion-exchange softening achieves Ca²⁺/Mg²⁺ removal through cationic substitution. After intermediate storage, water passes through cartridge filters (5 µm) to protect downstream equipment. The RO array provides primary desalination via semipermeable membranes, achieving near-complete ionic rejection [10].

The multi-media filters, activated carbon units, and softening resin are regenerable. Multi-media filters and activated carbon filters undergo hydraulic backwashing to restore capacity, while softening resins are regenerated with NaCl brine to displace accumulated ions and restore exchange capacity, ensuring sustainable long-term operation. Product water from the aboveground system is transferred underground through a 1,300 m stainless-steel pipeline. With an elevation differential of approximately 450 m, water first passes through a pressure-reducing valve before entering the storage tank.

The underground water system includes secondary RO membranes (enhanced desalination), an ElectroDeIonization (EDI) device, cartridge filters (0.1 µm pore size), Total Organic Carbon (TOC) removal, Ultra-Violet (UV) sterilization, degassing membranes, a microbubble generator, heat exchangers, and Ultra-Filtration (UF). The secondary RO membranes provide additional ion removal. EDI units combine ion-selective membranes and resin under DC fields for continuous deionization without chemical regeneration [11]. Cartridge filters capture residual particles. TOC/UV systems sterilize through UV irradiation while oxidatively decomposing organics. Degassing membranes, employing hollow fibers, vacuum pumps, and nitrogen purging, remove dissolved gases via partial pressure differentials and are critical for ²²²Rn removal from water [12–14]. Microbubble generators enhance the ²²²Rn removal efficiency of degassing membranes by reloading nitrogen into water [12, 14]. Heat exchangers regulate water temperature. UF membranes (≤10 nm pore size) remove trace amounts of ions, particles, colloids, and bacteria from water.

The 100 t/h UPW system supplies water for both the WCD and the initial filling of the CD. Owing to their distinct requirements, the two streams undergo separate treatments. The CD-bound water, which contacts the LS directly, receives additional UF for enhanced removal of ionic contaminants (²³⁸U/²³²Th/²²⁶Ra). The WCD-bound water, however, is routed through a dedicated heat exchanger to remove heat from the submerged PMT electronics, thereby ensuring stable detector temperatures.

²²⁶Ra Concentration in Water Measurement

The required ²²⁶Ra concentration of 50 µBq/m³ (∼2 × 10⁻²¹ g/g) is beyond the detection capability of conventional methods like high-purity germanium spectrometry [15] and inductively coupled plasma mass spectrometry [16]. Consequently, we developed a custom system that determines ²²⁶Ra activity by measuring its gaseous daughter, ²²²Rn.

Measurement Principle

The initial step in measuring ²²⁶Ra concentration in water via ²²²Rn emanation is ²²⁶Ra pre-concentration. The method leverages the strong adsorption of ²²⁶Ra onto MnO₂ [7, 17, 18], using laboratory-synthesized manganese fiber (Mn-fiber) [19, 20]. Following ²²⁶Ra extraction, the activity of ²²²Rn emanated from the Mn-fiber during a sealed period is measured. [FIGURE:3] illustrates the relevant decay branch of ²²⁶Ra. During sealing, the relationship between the activity of ²²⁶Ra and its ²²²Rn progeny activity can be calculated according to Eq. 1:

$$A_{Rn}(t) = A_{Ra} \times \frac{\lambda_{Rn}}{\lambda_{Rn} - \lambda_{Ra}} \times (e^{-\lambda_{Ra}t} - e^{-\lambda_{Rn}t})$$

where $A_{Rn}(t)$ is the ²²²Rn activity at time $t$ in mBq, $A_{Ra}$ is the initial activity of ²²⁶Ra extracted by Mn-fiber in mBq, $\lambda_{Rn}$ is the decay constant of ²²²Rn, $\lambda_{Ra}$ is the decay constant of ²²⁶Ra, and $t$ is the Mn-fiber sealing time in days.

Given that the half-life of ²²⁶Ra is significantly longer than that of ²²²Rn (i.e., $\lambda_{Ra} \ll \lambda_{Rn}$), Eq. 1 can be simplified to:

$$A_{Ra} = \frac{A_{Rn}(t)}{1 - e^{-\lambda_{Rn}t}}$$

Measurement Procedures and Devices

The ²²⁶Ra concentration measurement comprises four steps: (1) Mn-fiber synthesis, (2) ²²⁶Ra extraction from water, (3) Mn-fiber encapsulation, and (4) ²²²Rn activity determination.

1. Mn-fiber

To meet JUNO's requirement for µBq/m³ level ²²⁶Ra measurements, we significantly reduced the ²²⁶Ra background of the key adsorbent, Mn-fiber, from ∼17 µBq/g to 10.7 ± 1.5 µBq/g. This was achieved through three major enhancements to the standard synthesis protocol [19, 20]: (1) nitric acid pre-cleaning of the polyurethane fibers, (2) moving the boiling and rinsing processes to a cleanroom, and (3) substituting distilled water with high-purity water from China Resources C'estbon Beverage (China) Co., Ltd for rinsing—the most impactful change.

The Mn-fiber itself is produced by creating a firm MnO₂ coating on polyurethane fibers via reaction of KMnO₄ with H₂SO₄ under heated conditions:

$$2 \text{KMnO}_4 + 2 \text{H}_2\text{SO}_4 \xrightarrow{\Delta} \text{Mn}_2\text{O}_7 + 2 \text{KHSO}_4 + \text{H}_2\text{O}$$

$$\text{Mn}_2\text{O}_7 \rightarrow 4 \text{MnO}_2 + 3 \text{O}_2 \uparrow$$

[FIGURE:4] shows a picture of the laboratory-synthesized Mn-fiber. Gravimetric analysis indicates a ∼13% mass increase in polyurethane fibers after conversion to Mn-fiber.

2. Extraction Column

The ²²⁶Ra extraction column, illustrated in [FIGURE:5], is constructed from an acrylic tube (∼20 mm inner diameter and ∼185 mm length) sealed with rubber gaskets and 10 mm quick-connects. The assembled unit withstands water pressures up to ∼5 kg/cm². Approximately 5 g of Mn-fiber is packed in the central section, flanked by polyurethane fibers at both ends to filter particulate matter. During operation, water is directed upward through the column to maximize contact with the Mn-fiber. A residential water meter (GB/T 778 compliant [21], ±2% accuracy) at the outlet measures the total volume of water passing through the Mn-fiber.

3. ²²²Rn Emanation Measurement System

The ²²⁶Ra activity extracted by Mn-fibers was quantified by measuring its gaseous daughter ²²²Rn using a radon emanation measurement system ([FIGURE:6]). The system comprises a radon detector, a Mn-fiber container, dehumidification systems, a mass flow controller (MFC, Model 1179A, MKS), and a vacuum pump (Model ACP40, Pfeiffer Vacuum). All connections use metal-sealed ConFlat flanges and VCR components, ensuring leak tightness <1 × 10⁻⁹ Pa·m³/s. Key components are summarized below, with full details described in our previous work [20].

(1) Radon detector. The radon detector operates on the principle of electrostatic collection [22–28], where detection efficiency for radon progeny correlates with applied collection voltage and gas humidity. At 3% relative humidity and -700 V collection voltage, the detector achieves 90% collection efficiency for ²²²Rn progeny ²¹⁴Po. This corresponds to a Calibration Factor (CF) of 67 ± 6.7 counts per hour per Bq/m³ (cph/(Bq/m³)). The system's single-day measurement sensitivity is 0.7 mBq/m³ for ²²²Rn [22]. For each sample measurement, the radon detector integrates radon activity over a 24-hour data-taking period. As the measured value represents the average ²²²Rn concentration during this interval, determination of the initial ²²²Rn concentration requires correction for radioactive decay, calculated according to Eq. 5:

$$A_{Rn} = A_{Rn-M} \times \frac{\lambda_{Rn} \times (t_2 - t_1)}{e^{-\lambda_{Rn}t_1} - e^{-\lambda_{Rn}t_2}}$$

where $A_{Rn}$ is the initial ²²²Rn activity after transferring into the detector in mBq, $A_{Rn-M}$ is the measured ²²²Rn activity in mBq, $t_1$ is the start measurement time in hours, and $t_2$ is the end measurement time in hours.

(2) Mn-fiber container. Two chamber types are used ([FIGURE:7]): a standard CF35 tube (∼150 mm length) for 5 g samples, and a custom chamber (∼130 mm diameter and ∼300 mm length) for ≤300 g samples background measurement. All inner surfaces were electropolished to a roughness of <0.1 µm. After loading, chambers are purged with boil-off nitrogen to remove residual air and sealed for several days to allow ²²²Rn ingrowth. The activity ratio between ²²⁶Ra and ²²²Rn can be calculated according to Eq. 2. For instance, after a 10-day sealing interval, ²²²Rn activity reaches 83.7% of the initial ²²⁶Ra activity.

(3) Dehumidification systems. Mn-fibers that extracted ²²⁶Ra retain substantial water, releasing significant water vapor during radon transfer that degrades detector efficiency through humidity interference. This system comprises a liquid nitrogen tank, a solenoid valve, a dewar, a PT100 temperature sensor, and a temperature controller. Through thermostatic regulation of intermittent liquid nitrogen injection into the dewar, the system maintains -60 °C inside the dewar, reducing relative humidity to 3% at a 1 L/min gas flow rate without introducing background.

(4) MFC. The MFC regulates gas flow rate during radon transfer operations, with an adjustment range of 0.5–5 L/min.

(5) Vacuum pump. Before measurements, the pump evacuates the radon detector and associated pipelines. Subsequently, boil-off nitrogen purges the emanated ²²²Rn from the Mn-fiber container into the radon detector.

(6) Pipelines and valves. All stainless steel gas pipelines and interconnecting valves were EP-grade components.

Extraction Efficiency Calibration

The efficiency of extracting ²²⁶Ra from water by Mn-fiber was calibrated using a standardized ²²⁶Ra solution with a concentration of (9.04 ± 0.53) × 10³ µBq/m³. Our prior work [20] demonstrates consistent extraction efficiency across ²²⁶Ra concentrations (760 – 6.1 × 10⁴ µBq/m³) and water flow rates (1.6 – 8 L/min). To cover ²²⁶Ra concentrations from 1 to 1 × 10⁴ µBq/m³ in JUNO, sample volumes are varied, requiring dedicated calibration of extraction efficiency for different volumes. Rather than absolute quantification, we employed a relative scheme with volume-matched control experiments to simultaneously mitigate detector systematic uncertainties and background.

[TABLE:1] presents the calibrated ²²⁶Ra extraction efficiencies across different water volumes. Throughout the test, an additional 150 µL of ²²⁶Ra solution was added per 300 L of water, and the water flow rate through the Mn-fiber was maintained at ∼5 L/min. For water volumes below 1 m³ passing through the Mn-fiber extraction column, ²²⁶Ra extraction efficiency approached ∼100%. Beyond this volume, a progressive decline in ²²⁶Ra adsorption efficiency was observed, likely caused by a small amount of adsorbed ²²⁶Ra MnO₂ shedding.

Sensitivity Estimation

The sensitivity of measuring ²²⁶Ra concentration in water at 95% confidence level can be estimated according to Eq. 6:

$$L = \frac{1.645 \times \sigma_{BG} \times V_s}{24 \times CF \times V_w \times \varepsilon} \times 10^{-6}$$

where $L$ is the sensitivity in µBq/m³, $\sigma_{BG}$ is the statistical uncertainty of the system's background event rate in Counts Per Day (CPD), $V_s = 0.042$ m³ is the total volume of the measurement system, $CF = 67 \pm 6.7$ cph/(Bq/m³) is the detector calibration factor, $V_w$ is the sample water volume in m³, and $\varepsilon$ is the ²²⁶Ra extraction efficiency listed in [TABLE:1]. This calculation assumes a normally distributed background event rate and radioactive equilibrium between ²²²Rn and ²²⁶Ra.

The system background was determined by measuring 5 g of unused Mn-fiber following the standard sealing and analysis procedure. Ten replicate measurements yielded a mean background of 2.7 CPD ([FIGURE:8]), reflecting contributions from both the apparatus and the Mn-fiber. Using the parameters above and Eq. 6, we derived a sensitivity of 3.9 µBq/m³ for ²²⁶Ra concentration in water. This calculation assumes 20 m³ of water passes through 5 g Mn-fiber with a ²²⁶Ra extraction efficiency of 89.2%, and the data-taking time for ²²²Rn activity measurement is 24 h.

Results

²²⁶Ra Concentrations Across the System

Using the custom-developed device described above, we measured ²²⁶Ra concentrations in water at different locations of the JUNO 100 t/h UPW system, marked as S1–S17 in [FIGURE:2]. The results are shown in [TABLE:2]. The water volume for S1–S5 was 0.5 m³, for S6–S8 and S10 was 1 m³, for S9 and S11–S14 was 5 m³, and for S15–S17 was 20 m³. All measurements were performed after a 10-day Mn-fiber sealing period and a 24-hour ²²²Rn counting interval. Reported uncertainties include both statistical and systematic contributions from extraction efficiency. Sample S17 registered only 2 counts during measurement, below the mean detector background, and thus only an upper limit is provided.

In low-salinity solutions, ²²⁶Ra exists predominantly as uncomplexed Ra²⁺ [29–31], making it removable via deionization processes. According to [TABLE:2], this system reduces ²²⁶Ra concentration from ∼4 × 10⁵ µBq/m³ to <4 µBq/m³, representing a reduction of over five orders of magnitude. This >99.99% removal efficiency was achieved primarily through the synergistic action of softening resin and two-stage RO membranes, with each exhibiting >99% ²²⁶Ra removal efficiency. The ²²⁶Ra removal efficiencies of the main devices are shown in [TABLE:3]. The slight variation in removal efficiency between the two RO stages is primarily attributable to influent ²²⁶Ra concentration differences. This system employs two polishing resins in series, with tabulated efficiency calculations assuming identical ²²⁶Ra removal performance.

However, according to the data in [TABLE:2], mechanical filtration units (e.g., filter bags, cartridges, multi-media filters) also showed incidental ²²⁶Ra removal via adsorption of Ra²⁺ onto particulates, enabling co-removal during filtration. Among all devices, only RO integrates both deionization and particulate removal [32]. Additionally, the data indicate that ²²⁶Ra concentration in water increased after passing through each tank, primarily attributable to air ingress through tank ventilation valves. Fiberglass-reinforced plastic tanks require pressure-equalizing vent valves to prevent structural damage during water level fluctuations. These valves introduce ambient air (unfiltered for aboveground tanks, filtered for underground tanks) containing dust-bound ²²⁶Ra progenitors that solubilize upon water contact. A significant increase in ²²⁶Ra concentration after the pre-treatment water tank was traced to sediment introduced during initial commissioning when sea salt was used for softening resin regeneration. Although sea salt was replaced by clean regeneration salt and the tank was flushed, residual sediment on tank walls persisted as a leaching source.

Relationship Between ²²⁶Ra Concentration, Particulate Counts, and Resistivity

[TABLE:2] also lists particulate counts and resistivity values across the water system. Particulate counts were measured using a Bettersize C400 optical particle counter (Dandong Baxter); the small sample volume (20 mL) contributed to significant data uncertainties. Resistivity was measured using an AZ8302 portable conductivity tester (Taiwan Hengxin) upstream of the secondary RO and a +GF+ Signet 9900 in-line meter downstream. Sampling measurements become unreliable when resistivity exceeds 10 MΩ·cm, and in-line measurements prevent modification of the sampling point. These constraints resulted in missing high-resistivity data at multiple locations.

The apparent correlations between ²²⁶Ra concentration, particulate counts, and resistivity are merely a coincidental outcome of the purification process. The softening resin achieves >99% ²²⁶Ra removal by ion exchange (Na⁺ replaces Ra²⁺), which minimally influences resistivity and does not filter particles due to its 100–500 µm bead size. The simultaneous decrease in all parameters is due to their collective removal through successive treatment stages.

Discussion

This system reduces ²²⁶Ra concentration in water from ∼4 × 10⁵ µBq/m³ to 4 µBq/m³. Even though RO provides high removal efficiency for particulate matter and ions, the softening resin plays a critical complementary role. Our resin screening revealed significant variation in ²²⁶Ra removal efficiency, with cationic resins demonstrating superior performance over mixed-bed (anionic/cationic) alternatives. Although the selected softening resin achieves high ²²⁶Ra rejection (>99%), its pre-RO positioning is mandatory and non-interchangeable with post-RO polishing resin due to substantial organic leaching. These leachates cause measurable deterioration of water quality, primarily through resistivity reduction and TOC elevation.

Two tasks have been identified for reducing ²²⁶Ra pollution sources: (1) Thoroughly clean the pre-treatment water tank. Periodic regeneration of softening resin, commonly performed with economical sea salt, was found to introduce sediment containing significant ²²⁶Ra contamination. During the initial commissioning phase, conventional sea salt caused a sharp increase in ²²⁶Ra concentration—from ∼500 µBq/m³ to ∼3 × 10⁵ µBq/m³—after water passed through the pretreatment tank. Visual inspection revealed a substantial sediment layer at the tank bottom, which subsequent analysis confirmed originated from the sea salt. The data presented in this paper were collected after switching to clean regenerant salt and performing systematic flushing. However, residual sediment remains adhered to tank walls, explaining the persistent increase in ²²⁶Ra concentration observed after the pretreatment water tank. To fully eliminate this contamination source, thorough cleaning of the tank interior is required. As this procedure entails tank drainage and temporary system shutdown, it will be scheduled for a suitable maintenance window.

(2) Mitigate contamination from storage tank breather valves. Storage tanks introduce measurable radiological contamination primarily through their air-admittance breather valves. Two mitigation strategies have been evaluated: installing particulate-retentive filters on breather assemblies or using boil-off nitrogen as a clean breathing gas. While nitrogen blanketing could thoroughly eliminate background contamination, it would require a dedicated gas supply system at significant cost. Furthermore, existing boil-off nitrogen capacity at the JUNO site is insufficient to meet operational demand. Therefore, installation of filtration systems at each tank air inlet has been selected as the practical solution, which will be implemented during a suitable maintenance period.

Summary

JUNO is a multi-purpose neutrino experiment primarily designed to determine the neutrino mass ordering. To ensure detector safety and minimize background, the CD must be filled with UPW containing <50 µBq/m³ of ²²⁶Ra before LS introduction. A dedicated 100 t/h UPW production and circulation system was developed, integrating multi-stage purification—including specialized softening resin and reverse osmosis—with a novel, highly sensitive (3.9 µBq/m³) ²²⁶Ra measurement system. Measurement results across purification stages demonstrated >99% ²²⁶Ra removal efficiency for both the softening resin and RO. Key contamination pathways, including water tank venting and regenerant salt impurities, were identified, and corresponding mitigation strategies have been proposed.

The JUNO 100 t/h water system successfully reduced ²²⁶Ra concentration from ∼4 × 10⁵ µBq/m³ to <4 µBq/m³, achieving a reduction exceeding five orders of magnitude. This performance not only satisfies JUNO's stringent technical requirements but also represents a world-leading achievement in ultra-low-²²⁶Ra water purification.

References

[1] F. An, G. An, Q. An et al., Neutrino physics with JUNO. J. Phys. G 43, 030401 (2016). https://doi.org/10.1088/0954-3899/43/3/030401
[2] JUNO Collaboration, JUNO physics detector. Prog. Part. Nucl. Phys. (2022). https://doi.org/10.1016/j.ppnp.2021.103927
[3] C. Cao, J. Xu, M. He et al., Mass production and characterization of 3-inch PMTs for the JUNO experiment. Nucl. Instrum. Meth. Phys. Res. Sect. A 1005, 165347 (2021). https://doi.org/10.1016/j.nima.2021.165347
[4] JUNO Collaboration, Radioactivity control strategy for the JUNO detector. J. High Energy Phys. 2021, 102 (2021). http://doi.org/10.1007/JHEP11(2021)102
[5] SuperK Collaboration, The Super-Kamiokande Detector. Nucl. Instrum. Meth. Phys. Res. Sect. A 501, 418-462 (2003). https://doi.org/10.1016/S0168-9002(03)00425-X
[6] SNO Collaboration, The Sudbury Neutrino Observatory. Nucl. Instrum. Meth. Phys. Res. Sect. A 449, 172-207 (2000). https://doi.org/10.1016/S0168-9002(99)01469-2
[7] T.C. Andersen, I. Blevis, J. Boger et al., Measurement of radium concentration in water with Mn-coated beads at the Sudbury Neutrino Observatory. Nucl. Instrum. Meth. Phys. Res. Sect. A 501, 399-417 (2003). https://doi.org/10.1016/S0168-9002(03)00616-8
[8] T. Ueno, Radium Concentration Measurements in Super-Kamiokande. Master's Thesis, University of Tokyo (2021). https://www-sk.icrr.u-tokyo.ac.jp/assets/paper/list/pdf/mthesis/2008/master-ueno.pdf
[9] Electronic Grade Ultrapure Water, SJ/T 11452-2013 (2013). https://std.samr.gov.cn/
[10] M. Tawalbeh, L. Qalyoubi, A. Al-Othman et al., Insights on the development of enhanced antifouling reverse osmosis membranes: Industrial applications and challenges. Desalination 553, 116460 (2023). https://doi.org/10.1016/j.desal.2023.116460
[11] Z.U. Khan, M. Moronshing, M. Shestakova et al., Electro-deionization (EDI) technology for enhanced water treatment and desalination: A review. Desalination 548, 116254 (2023). https://doi.org/10.1016/j.desal.2022.116254
[12] B. Wang, C. Guo, Y.G. Xie et al., Measurement of the radon concentration in the ultrapure water of JUNO-LS UPW system. J. Instrum. 20, P03028 (2025). https://doi.org/10.1088/1748-0221/20/03/P03028
[13] Y. Nakano, T. Hokama, M. Matsubara et al., Measurement of the radon concentration in purified water in the Super-Kamiokande IV detector. Nucl. Instrum. Meth. Phys. Res. Sect. A 977, 164297 (2020). https://doi.org/10.1016/j.nima.2020.164297
[14] T.Y. Guan, Y.P. Zhang, B. Wang et al., Development of low-radon ultra-pure water for the Jiangmen Underground Neutrino Observatory. Nucl. Instrum. Meth. Phys. Res. Sect. A 1063, 169244 (2024). https://doi.org/10.1016/j.nima.2024.169244
[15] G.R. Araujo, L. Baudis, Y. Biondi et al., The upgraded low-background germanium counting facility Gator for high-sensitivity γ-ray spectrometry. J. Instrum. 17, P08010 (2022). https://doi.org/10.1088/1748-0221/17/08/P08010
[16] S. Ito, Y. Takaku, M. Ikeda et al., Determination of trace level thorium and uranium in high purity gadolinium sulfate using ICP-MS with solid-phase chromatographic extraction resins. AIP Conf. Proc. 1921, 030003 (2018). https://doi.org/10.1063/1.5018990
[17] W.S. Moore, D.F. Reid, Extraction of radium from natural waters using manganese-impregnated acrylic fibers. J. Geophys. Res. 78, 8880-8886 (1973). https://doi.org/10.1029/JC078i036p08880
[18] R.N. Peterson, W.C. Burnett, N. Dimova et al., Comparison of measurement methods for radium-226 on manganese-fiber. Limnol. Oceanogr. 7, 196-205 (2009). https://doi.org/10.4319/lom.2009.7.196
[19] L.F. Xie, J.C. Liu, S.K. Qiu et al., Developing the radium measurement system for the water Cherenkov detector of the Jiangmen Underground Neutrino Observatory. Nucl. Instrum. Meth. Phys. Res. Sect. A 976, 164266 (2020). https://doi.org/10.1016/j.nima.2020.164266
[20] C. Li, B. Wang, Y. Liu et al., Developing a µBq/m³ level ²²⁶Ra concentration in water measurement system for the Jiangmen Underground Neutrino Observatory. Nucl. Instrum. Meth. Phys. Res. Sect. A 1063, 169257 (2024). https://doi.org/10.1016/j.nima.2024.169257
[21] China National Standards Full-Text Disclosure System. https://openstd.samr.gov.cn
[22] J.X. Wang, T.C. Andersen, J.J. Simpson, An electrostatic radon detector designed for water radioactivity measurements. Nucl. Instrum. Meth. Phys. Res. Sect. A 421, 601-609 (1999). https://doi.org/10.1016/S0168-9002(98)01230-3
[23] Y.Y. Chen, Y.P. Zhang, Y. Liu et al., A study on the radon removal performance of low background activated carbon. J. Instrum. 17, P02003 (2022). https://doi.org/10.1088/1748-0221/17/02/P02003
[24] C. Guo, J.C. Liu, Y.P. Zhang et al., Study on the radon removal for the water system of Jiangmen Underground Neutrino Observatory. Radiat. Detect. Technol. Methods 2, 48 (2018). https://doi.org/10.1007/s41605-018-0077-8
[25] Y. Liu, Y.P. Zhang, J.C. Liu et al., System upgrade for µBq/m³ level ²²²Rn concentration measurement. J. Instrum. 18, T03002 (2023). https://doi.org/10.1088/1748-0221/18/03/T03002
[26] C. Li, Y. Zhang, L. Lv et al., Study on the radon adsorption capability of low-background activated carbon. J. Radioanal. Nucl. Chem. 333, 337-346 (2024). https://doi.org/10.1007/s10967-023-09211-w
[27] Y. Nakano, H. Sekiya, S. Tasaka et al., Measurement of radon concentration in super-Kamiokande's buffer gas. Nucl. Instrum. Meth. Phys. Res. Sect. A 867, 108-114 (2017). https://doi.org/10.1016/j.nima.2017.04.037
[28] Y. Nakano, T. Hokama, M. Matsubara et al., Measurement of the radon concentration in purified water in the Super-Kamiokande IV detector. Nucl. Instrum. Meth. Phys. Res. Sect. A 977, 164297 (2020). https://doi.org/10.1016/j.nima.2020.164297
[29] B.L. Dickson, Radium in groundwater. IAEA. Vol. 1, No. 1 (1990). https://inis.iaea.org/records/swprd-wbq70
[30] D. Porcelli and P.W. Swarzenski, The behavior of U- and Th-series nuclides in groundwater. Rev. Mineral. Geochem. 52, 317-361 (2003). https://doi.org/10.2113/0520317
[31] F. Carvalho, D. Chambers, S. Fernandes et al., The environmental behaviour of radium: Revised edition. IAEA Technical Reports Series No. 476 (2014). http://www.iaea.org/books
[32] PURETEC Industrial Water, The basics - Reverse osmosis. https://puretecwater.com